add tests of population remapping

petrelharp commented 2 years ago

[x] alive nodes with population not having SLim metadata produces an error
[x] reading of multiple tree sequences whose populations conflict produces an error
[x] reading of multiple tree sequences whose populations are remapped to not conflict works
[x] existing non-SLiM metadata in population table entries not referred to by alive nodes is not modified (including at the end of the table, past all slim populations)
[x] saving out of a tree sequence with an empty pop in nonWF still writes out the empty pop, and reloading it recovers that pop
[x] when remapping is used, carryover metadata gets shifted to the new remapped index unmodified, without errors.
[x] a remapping that leads to a conflict between the remapped subpops correctly errors out
[x] when carryover metadata is present, SLiM correctly requires a remapping entry for that subpop index (if any remapping is being done at all)
[x] when a migration table is present, which SLiM does read in for WF models, remapping of the source_subpop for the migrations works correctly (which it definitely does not at the moment)

petrelharp commented 2 years ago

So, we're testing, after loading a ts into SLiM and writing it back out again, that:

an empty pop in WF throws an error
an empty pop in nonWF does not throw an error
names are preserved
non-SLiM metadata is preserved (including at the end of the table, past all slim populations)
alive nodes with population not having SLiM metadata produces an error
remapping the same input pop to two different subpops errors
and (not yet working):
remapping with mixed SLiM and non-SLiM populations works and retains non-SLiM pops
that migration records in subpop metadata are correctly remapped-

petrelharp commented 2 years ago

In this method we test that remapping is correct for:

tskit-dev / pyslim